Overview
Mistral AI is a French model maker. Its API, managed in Mistral Studio, serves the Mistral Large, Medium, Small and Ministral text models, Codestral for code, OCR for documents and Voxtral for speech, plus a hosted third-party model (GLM 5.2). Most of its models carry an 'Open' label, meaning the weights are published, so a workload can later move to self-hosting.
Prices are low for the model sizes on offer: Mistral Large 3 costs $0.50 in and $1.50 out per million tokens and Small 4 costs $0.15 / $0.60. Batch processing halves prices, cached input tokens are 90% cheaper, and Priority processing is available for predictable fast compute. Choosing an EU or global regional endpoint adds 10%.
The consumer Free plan includes $10 a month of API credits, enough to prototype. Note the data setting before sending real data: Mistral says API calls may be used to improve its services unless you switch off the 'Anonymous improvement data' toggle, which is separate from the chat-app opt-out. Enterprise APIs with SLAs and higher limits cost 75% above list.
Pricing and plans
Free tier
The Free plan includes $10 a month in API credits for Mistral Studio; exact rate limits are shown in the account's Limits page.
At the limit Not stated The vendor does not say what happens at the limit, or we have not found it yet.
| Plan | Price | What you get |
|---|---|---|
| Mistral Large 3 | $0.50 in / $1.50 out per 1M | Open-weight flagship multimodal model |
| Mistral Medium 3.5 | $1.50 in / $7.50 out per 1M | Open; agentic and coding use |
| Mistral Small 4 | $0.15 in / $0.60 out per 1M | Apache 2.0 |
| Codestral | $0.30 in / $0.90 out per 1M | Code completion and fill-in-the-middle |
| Ministral 3 (3B / 8B / 14B) | $0.10 / $0.15 / $0.20 per 1M | Same price for input and output |
| OCR 4.1 | $4 per 1,000 pages | Document AI $5 per 1,000 pages |
| Batch | 50% off | Cached input is 90% off; EU or global regional inference +10% |
| Enterprise APIs | List + 75% | Regional processing controls, SLAs, higher rate limits, premium support |
What it does
- Open-weight models
- Mistral Large 3, Medium 3.5, Small 4 (Apache 2.0) and Ministral 3 are marked Open on the pricing page.
- Regional inference
- Global or EU inference endpoints for supported models, at +10%.
- Batch and caching
- Batch at half price; cached input tokens 90% cheaper.
- Document OCR
- OCR 4.1 at $4 per 1,000 pages, Document AI at $5 per 1,000 pages.
- Speech
- Voxtral transcription from $0.003 per audio minute and Voxtral TTS at $0.016 per 1,000 characters.
- Fine-tuned classifiers
- Train Ministral-based classifiers from $1 per 1M training tokens plus $2 per model per month storage.
Strengths and limitations
Strengths
- EU company with an EU inference endpoint option.
- Most models have published weights, so there is an exit path to self-hosting.
- Low list prices, e.g. Mistral Large 3 at $0.50 / $1.50 per 1M tokens.
- A monthly free API credit rather than a one-off trial.
Limitations
- API data may be used to improve Mistral's services until you switch off a separate opt-out toggle.
- Commercial self-hosting of some open-weight models requires a Mistral licence, per its FAQ.
- Rate limits are not published on the pricing page; they appear in the account's Limits page.
- Enterprise API features (SLAs, higher limits) cost 75% above list.
Who it suits
Good for
- European teams that want EU inference
- Document processing with OCR
- Cost-sensitive text workloads that may move to self-hosted weights later
Look elsewhere if
- Teams that need OpenAI, Anthropic or Google models behind the same key
- Anyone who cannot accept an opt-out (rather than opt-in) data setting
Alternatives to Mistral API (Mistral Studio)
| Tool | From | Free option | Stages |
|---|---|---|---|
| OpenRouter Model router (unified API) | — | Free tier | |
| SambaNova Cloud (SambaCloud) Inference provider (open-weight models) | Usage-based | Free tier | |
| Scaleway Generative APIs Inference provider (EU-hosted open-weight models) | Usage-based | ||
| Cerebras Inference Inference provider (open-weight models) | Usage-based | ||
| GroqCloud Inference provider (open-weight models) | Usage-based | Free tier | |
| Fireworks AI Inference provider (open-weight models) | Usage-based |
Questions
Does Mistral have a free API tier?
Mistral's Free plan includes $10 a month in API credits for Mistral Studio, according to its pricing page on 2026-09-29.
Does Mistral train on API data?
Mistral's help centre says API calls and related data may be used to improve its services unless you disable the 'Anonymous improvement data' toggle in the Admin panel.
Can I run Mistral models in the EU?
Yes. Mistral offers global or EU inference endpoints for supported models at a 10% premium.
How much does Mistral Large 3 cost?
$0.50 per 1M input tokens and $1.50 per 1M output tokens on 2026-09-29, half that through Batch.
Sources read for this page